Khant S.

I'm an AI engineer who builds AI products end to end: the agent systems, the multi-tenant SaaS around them, the public APIs, and the cloud infrastructure that runs it all in production. For the past two years I led engineering across five AI products at DASOMX (South Korea), including two multi-tenant SaaS platforms I built end to end. OKCHAT: AI chatbot and voice agent platform (1,000+ organizations) • Self-serve chatbot builder with knowledge bases from documents, Q&A pairs, and CSVs, powered by hybrid RAG with reranking • Omnichannel deployment: embeddable website widget, Kakao, LINE, Naver TalkTalk, Slack, and Discord • AI voice agents, including phone calling through telephony integration • Custom API tools and MCP server support, so agents can take action in external systems • Support for all major frontier model providers (Anthropic, OpenAI, Google Gemini, DeepSeek, and more) with routing and fallback, plus local models via Ollama • Public REST API for external developers: chatbots, knowledge bases, chat, conversations, and voice agents LEIA: AI learning platform (1.5K+ courses) • Source-grounded generation of courses, quizzes, and podcasts from uploaded material • AI video generation: I built the full orchestration, from AI-generated slide decks to narrated voice-over • AI tutors with real-time voice • Public API for external users • Distributed backend with job queues, async workers, and workflow orchestration, running in separate dev, staging, and production environments on Azure DevOps and infrastructure • Infrastructure as code with Terraform: I set up LEIA's dev, staging, and production environments on Azure, each running multiple services • Containerized services on Azure Container Apps, with KEDA autoscaling for queue-driven workers • CI/CD with GitHub Actions for automated, repeatable deployments across environments • Observability with OpenTelemetry tracing from API requests through to background jobs • Production experience on both AWS and Azure What I build for businesses End-to-end agent systems, from first prototype to production. I design the agent and its tools, connect it to your data and systems, build the evaluations that prove it works, and deploy and operate it at scale. That includes: • Agents that take real action through tools, APIs, and the software you already use • Knowledge retrieval over your documents, measured with evals before and after every change • Low-latency voice agents for support, sales, and training • The multi-tenant backend, public API, and full-stack product around the agent • Cloud infrastructure, CI/CD, and monitoring so it runs reliably in production • Cost control through multi-model routing and fallback I'm framework-agnostic and choose the tools that fit the problem, not the other way around. Stack TypeScript, Python, Go · Next.js, React, Node.js, FastAPI · Vercel AI SDK, Mastra, LiveKit (voice agents) · PostgreSQL, pgvector, Pinecone, Redis, RabbitMQ, BullMQ · AWS, Azure, Terraform, Docker, Coolify, GitHub Actions, OpenTelemetry If you have a production AI problem around cost, accuracy, latency, or scale, send me the details and I'll tell you how I'd approach it.